washdry.Rmd

---
title: "WashDry"
author: "Ryszard Czerminski"
date: "2017-09-19"
header-includes:
   - \usepackage{amsmath}
   - \usepackage{color}
   - \usepackage{xcolor}
   - \usepackage{mathtools}
   - \DeclareMathOperator*{\E}{\mathbb{E}}
   - \usepackage{hyperref}
   - \hypersetup{
      colorlinks=true,
      linkcolor=blue,
      filecolor=magenta,
      urlcolor=cyan,
    }

output:
    pdf_document
---

\section{Intro}

\section{Setup global contants for data and libs}
```{r}
cat('cwd:', getwd(), '\n')
dir <- list()
dir$libs <- Sys.getenv('SPRINGBOARD_LIBS')
dir$data <- Sys.getenv('SPRINGBOARD_DATA')
fn.libs <- paste(dir$libs, 'sigUtils.R', sep='/')
probs <- seq(0.1, 0.9, 0.1)
probs <- seq(0.25, 0.75, 0.5)
cat('SPRINGBOARD_LIBS:', dir$libs, '\n')
cat('SPRINGBOARD_DATA:', dir$data, '\n')
cat('fn.libs:', fn.libs, '\n')
```

\section{Production line sample}

```{r}
fn <- paste(dir$data, 'sample.csv', sep='/')
x <- read.csv(fn)
fmt <- '%m/%d/%Y %I:%M %p'
x$Start <- as.POSIXct(x$Start.Time, format=fmt)
x$End <- as.POSIXct(x$End.Time, format=fmt)
x$Start.Time <- x$End.Time <- NULL
x$Span <- x$End - x$Start
```
Note, that \texttt{Duration} (from csv) and \texttt{Span} (calculated as diff beteen \texttt{Start}
and \texttt{End} are not the same:

```{r}
head(x[, c('Duration', 'Span')])
head(x)
```

I am removing \texttt{Duration} column and will use \texttt{Span} column for know.
`r x$Duration <- NULL`
\textbf{You should find out what is the reason for this discrepancy.}
```{r}
head(x)
```
\subsection{Are there any spans of length 0?}
```{r}
w <- which(x$Span == 0)
x[w, c('State', 'Reason', 'Span')]
```
Apparently there are. Let's remove them since these do not seem
to be very meaningful. I will remove old row numbering as well.
```{r}
x <- x[-w,]
row.names(x) <- NULL
```

\subsection{Check for gaps}
```{r}
i <- 1:(nrow(x)-1); j <- i + 1
d <- x$Start[j] - x$End[i]
w <- which(d > 0)
```
There is a `r d[w]` `r units(d[w])` gap in data around row `r w`.
Code needs to be able to deal with time gaps like this.

\subsection{How many different states we have?}

```{r}
tb <- table(x$State)
tb
```
Let's recode them as follows
```{r}
codes <- c(Running=0, Standby=1, Setup=2, Down=3)
```
Now we can convert the data frame. Let's just select a fixed slice of a df
e.g.8h beetween 8am and 4pm)
```{r}
slice <- list()
slice$start <- as.POSIXct('2017-08-11 08:00:00', format='%Y-%m-%d %H:%M:%S')
slice$end <- as.POSIXct('2017-08-11 16:00:00', format='%Y-%m-%d %H:%M:%S')
```


\subsection{Read vib data}
```{r}
source(fn.libs)
data.sets <- c('dry', 'wash')
cols <- c('red', 'blue')
names(cols) <- data.sets
v <- prep.signals(data.sets, dir=dir$data, verbose=1)
if (length(v$err) > 0) {
  for (x in v$err) cat(x, '\n')
  stop()
} else {
  signal <- v$data
}

```
```{r}
plt.signals(signal)
```

\section{Signal amplitude analysis}

```{r}
plt.cdfs(signal, xlab='amplitude', value='norm', titl='signals')
qSig <- list() # list of signal quantiles
qSig$dry <- quantile(signal$dry$norm, probs)
qSig$wash <- quantile(signal$wash$norm, probs)
```

\section{Frequency analysis}
```{r}
source(fn.libs)
power <- list()
for (name in names(signal)) {
  y <- signal[[name]]$norm; t <- signal[[name]]$t
  v <- get.fft(y, t, tag=name)
  plt.power(y, t, v$pw, v$f, tag=name)
  power[[name]] <- list(f=v$f, pw=v$pw)
}
```
```{r}
qPow <- list() # power quantiles
plot(1, xlim=c(0,50), ylim=c(0,1), type='n', xlab='freq', ylab='Cummulative power')
for (x in names(power)) {
   f <- power[[x]]$f; p <- power[[x]]$pw
   v <- cumsum(p)
   v <- v/max(v)
   qf <- splinefun(v, f)
   qPow[[x]] <- qf(probs)
   lines(f, v, type='l', col=cols[x])
}
legend('topleft', legend=names(power), col=cols, lwd=2)
```
\section{Quantiles for commulative power spectrum}
```{r}
cat('signal quantiles:\n')
print (qSig)
cat('power quantiles:\n')
print (qPow)
```